Skip to content
Abstract cover artwork for Feature Engineering That Actually Moves Metrics, a Machine Learning course Machine Learning
Machine Learning intermediate 11 hours 4.7

Feature Engineering That Actually Moves Metrics

Taught by Sofia Duarte

Most model gains come from features, not architecture. Target encoding without leakage, temporal features that respect causality, aggregation windows, interaction discovery and feature stores. Includes a rigorous ablation methodology so you can prove which features earned their place.

What you'll cover

feature engineering pandas encoding time series tabular

More in Machine Learning

Abstract cover artwork for Recommender Systems from Scratch, a Machine Learning course Machine Learning
Machine Learning intermediate

Recommender Systems from Scratch

Build a recommender end to end: implicit-feedback matrix factorisation, content-based retrieval with embeddings, two-stage retrieve-then-rank architectures, and the cold-start problem every real system faces. Evaluated with offline ranking metrics and an honest discussion of why they disagree with A/B tests.

HT Hiroshi Tanaka 4.9

$89.00

15 hours

→
Abstract cover artwork for Machine Learning Foundations with scikit-learn, a Machine Learning course Machine Learning
Machine Learning beginner

Machine Learning Foundations with scikit-learn

The honest introduction: how to frame a problem, build a baseline, and know when your model is fooling you. Linear and tree-based models, proper cross-validation, leakage detection, class imbalance, and calibration. Every concept is implemented on real tabular data rather than a toy dataset.

SD Sofia Duarte 4.8

$49.00

18 hours

→
Abstract cover artwork for Gradient Boosting in Depth: XGBoost and LightGBM, a Machine Learning course Machine Learning
Machine Learning intermediate

Gradient Boosting in Depth: XGBoost and LightGBM

Still the strongest default for tabular problems, and still widely misused. How the boosting objective actually works, which hyperparameters matter and which are noise, categorical handling, and reading SHAP values without over-reading them.

HT Hiroshi Tanaka 4.8

$75.00

11 hours

→
Abstract cover artwork for MLOps: Shipping Models That Survive Contact With Users, a Machine Learning course Machine Learning
Machine Learning advanced

MLOps: Shipping Models That Survive Contact With Users

Experiment tracking, model registries, reproducible training pipelines, shadow deployments, drift detection and automated rollback. You will build a pipeline where a retrained model cannot reach production without passing evaluation gates, and where a bad deploy is detected in minutes rather than quarters.

AR Ahmed Rahal 4.7

$119.00

20 hours

→